← Engel's Code Design Notebook

OpenHands agent-sdk verification

How agents check that the OpenHands Agent SDK does what it says, one feature at a time: a map of every behavior an agent can see, a CLI that drives the real thing, and known bugs kept as expected failures until someone fixes them. Everything here is open, and anyone can rerun it.

The explainer

Two minutes on what an agent actually sees when it uses the SDK: the map, the control-agent-server CLI, a recipe that caught a bug, and the full replay. Drawn with Manim, with a generated score. Download (MP4, 10 MB) · Scene script and music

Agent Server

The first map: 34 families and 931 sub-features of the Agent Server, driven over REST, WebSocket, the Python client and the TypeScript client. Its full replay: 771 pass, 156 known bugs reproduced, 6 blocked, no surprises. It lives in software-agent-sdk PR #5621.

A map that keeps itself honest
The presentation, opening with a short animated film: a control CLI, an executable map of 930 Agent Server behaviors, and known bugs as expected failures. The full replay, and the upstream fix it caught.
presentation + film · 11 October 2026
Agent Server bug tracker
38 Agent Server bugs the map found or reproduced: severity, status and a two-sentence summary each, linked to the issues and to umbrella #5653.
tracker · 9 October 2026

Core SDK

The next map covers the SDK used in process: the agent loop, LLMs, conversations, condensers, skills, MCP, tools, workspaces and more, in 21 families (S01–S21). It is being drafted now and will be linked here when it is published.

OpenHands app

The sister project: the same approach applied to the OpenHands app (Agent Canvas), with a browser-driving CLI and real before/after screenshots.

Give agents a way to prove it
A visual presentation of OpenHands’ verification loop: 716 mapped behaviors, a control CLI, evidence rules, and maintenance. Real before/after screenshots show what the first run found and fixed, with credit to Lauren Tan’s pstack.
presentation · updated 9 October 2026
120 findings: the issues and their fixes
Issues filed by OpenHands feature-map verification runs, grouped by repository and theme, with the checked status of their fixing pull requests.
findings snapshot
control-openhands: agents using Agent Canvas like a user
The CLI behind the verify-openhands skill: isolated stacks, a browser that persists across commands, and an evidence ledger.
how it works · 5 October 2026

OpenHands features

The feature map, for humans: 743 behaviors in 27 families, with readable CLI scripts and real screenshots. Source snapshot: 9 October 2026.

Explore the OpenHands feature map
Choose a family or search a behavior ID. Follow the user entry point, the commands, and the expected result.
openhands-features · full index

Starting work

F01 · First run, onboarding and sign-in
22 behaviors · 26 readable recipes and supporting checks.
openhands-features
F02 · App shell, sidebar and command menu
26 behaviors · 26 readable recipes and supporting checks.
openhands-features
F03 · Home and starting work
39 behaviors · 34 readable recipes and supporting checks.
openhands-features
F04 · Conversation list and folders
40 behaviors · 40 readable recipes and supporting checks.
openhands-features

Conversation & workspace

F05 · Composer, slash commands and plan mode
25 behaviors · 27 readable recipes and supporting checks.
openhands-features
F06 · Agent activity: messages, events and lifecycle
35 behaviors · 28 readable recipes and supporting checks.
openhands-features
F07 · Conversation page, header and management menu
25 behaviors · 26 readable recipes and supporting checks.
openhands-features
F08 · Workspace drawer: files and changes
33 behaviors · 31 readable recipes and supporting checks.
openhands-features
F27 · Workspace tools: terminal, browser, planner, tasks and usage
30 behaviors · 21 readable recipes and supporting checks.
openhands-features

Settings & configuration

F09 · Settings shell and navigation
25 behaviors · 21 readable recipes and supporting checks.
openhands-features
F10 · LLM profiles
34 behaviors · 33 readable recipes and supporting checks.
openhands-features
F11 · Provider connections
21 behaviors · 23 readable recipes and supporting checks.
openhands-features
F12 · Model router (meta-LLM profiles)
25 behaviors · 23 readable recipes and supporting checks.
openhands-features
F13 · Agent profiles
20 behaviors · 21 readable recipes and supporting checks.
openhands-features
F14 · Secrets
11 behaviors · 14 readable recipes and supporting checks.
openhands-features
F15 · Condenser, agent context and verification settings
23 behaviors · 24 readable recipes and supporting checks.
openhands-features
F16 · Application settings
19 behaviors · 21 readable recipes and supporting checks.
openhands-features

Customization

F17 · Customize hub and MCP servers
30 behaviors · 30 readable recipes and supporting checks.
openhands-features
F18 · Skills catalog
22 behaviors · 27 readable recipes and supporting checks.
openhands-features
F19 · Plugins and plugin launch
25 behaviors · 27 readable recipes and supporting checks.
openhands-features
F20 · Canvas apps
24 behaviors · 26 readable recipes and supporting checks.
openhands-features

Automations

F21 · Automations dashboard and actions
34 behaviors · 34 readable recipes and supporting checks.
openhands-features
F22 · Creating automations: templates, setup and import
29 behaviors · 29 readable recipes and supporting checks.
openhands-features
F23 · Automation detail, runs and editing
39 behaviors · 29 readable recipes and supporting checks.
openhands-features
F24 · Automation Git Sync
26 behaviors · 26 readable recipes and supporting checks.
openhands-features

Backends & runtimes

F25 · Backends, Cloud and sharing
29 behaviors · 27 readable recipes and supporting checks.
openhands-features
F26 · Launcher modes, Docker, desktop and library
32 behaviors · 28 readable recipes and supporting checks.
openhands-features